⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

Self-Supervised JEPA-based World Models for LiDAR Occupancy Completion and Forecasting

arXiv 26.2 2026 53.2 method

TLDR

Self-supervised JEPA-based world model for LiDAR occupancy completion and forecasting in autonomous driving.

Reasoning

The paper proposes a novel self-supervised world model using JEPA for spatiotemporal prediction from LiDAR data, with promising proof-of-concept results on occupancy completion and forecasting. However, the evaluation is limited to a single downstream task and lacks comparison to strong baselines or real-world deployment details.

Read-first score

Read-first score 53.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 30.

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%
70

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,experiment

Topical relevance 42%
42.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 25%
38

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 466.

Keyword Scores

world model
10
world dynamics prediction
8
video world model
4
world simulator
3
generative world model
2
model-based reinforcement learning world model
2
interactive world model
1

Deep Analysis

Innovations

  • Self-supervised JEPA-based world model for autonomous driving using LiDAR data
  • Joint-embedding predictive architecture (JEPA) applied to spatiotemporal evolution prediction from LiDAR
  • Downstream LiDAR occupancy completion and forecasting (OCF) task to evaluate learned representations

Methodology

The paper proposes AD-LiST-JEPA, a self-supervised world model that uses a joint-embedding predictive architecture (JEPA) to predict future spatiotemporal evolution from LiDAR data. The learned representations are evaluated through a downstream LiDAR-based occupancy completion and forecasting (OCF) task, which jointly assesses perception and prediction.

Key Results

Proof of concept experiments show better OCF performance with the pretrained encoder after JEPA-based world model learning compared to baselines.

Limitations

  • Proof of concept experiments only, limited scale and generalizability not yet validated

Resources

Paper

Tags